GifCities: the website that searches tiny moving history
12 Nov 2026 · By the Boring Websites editors
GifCities is the Internet Archive's search engine for animated GIFs preserved from GeoCities. Its small pictures offer a useful entrance to old personal websites, where a spinning ornament can be the beginning of a much larger visit.
In brief
- GifCities searches an archived collection rather than the whole present-day web.
- Its documentation distinguishes semantic search from searches of filenames and URL paths.
- Special Search includes exact width and height filters.
- The archive connection matters as much as the pictures themselves.
A search engine with a very small subject
The GifCities homepage gives the collection a plain description: GeoCities animated GIF search. Its suggested starting points include snowglobes, butterflies, balloons, sparkles and a drawing of a cat. That is a useful declaration of scale. The subject is not the entirety of digital culture. It is the little moving things that once helped a page look inhabited.
As an editorial proposition, this is unusually inviting. A visitor does not need to recognise an old domain or remember the address of a vanished homepage. A familiar object is enough. The search can begin with a butterfly, without requiring its owner, its original location or a reason for wanting it. An archive becomes approachable through something almost comically modest.
The suggested subjects also offer different ways of looking. A snowglobe suggests a bounded miniature. Sparkles suggest a texture or an effect. A cat suggests a recognisable figure, while a drawing of a cat adds a description of how that figure should appear. Those are proposed search exercises rather than promises about particular results. They give a casual visitor somewhere concrete to start.
A reasonable first visit can therefore have a small question. What would make a suitable little decoration for an imaginary page about cats? The question is deliberately narrower than a survey of internet history. It directs attention towards shape, movement and use. The visitor can compare objects rather than treating every result as the same undifferentiated substance called nostalgia.
The appeal is partly that such a search has an acceptable stopping point. One interesting object is enough. There is no editorial need to turn the visit into a complete inventory or to decide which image represents an entire era. A little curiosity suits a little image. The archive can accommodate a modest errand without making it feel like incomplete homework.
The archive behind the ornaments
The Internet Archive's original GifCities announcement, published on 1 November 2016, presents the project as a playful route into its web archive. It reports more than 4.5 million extracted animated GIFs, with 1.6 million unique images. Those are launch figures, not a fresh count of today's search results.
The distinction between extracted and unique is worth retaining. The two numbers describe different things. A collection can contain repeated appearances of an image without each appearance being a newly invented picture. The launch announcement makes that difference explicit. Quoting only the larger figure as a count of distinct creations would give the collection a tidier story than its source supports.
For a reader, an archive of repeated appearances can be interesting in its own right. A picture is an object; its placement on a page is another object of attention. The same ornament might suggest something different beside a dedication, a collection of links or an introduction. That is an interpretive reason to care about context, rather than a claim about any particular result.
The launch announcement says most GIFs linked to archived pages where they were displayed. That connection gives the search its best possible second act. The picture is a doorway. Instead of ending with an isolated moving object, a visit can become an examination of the page that accommodated it. The ornament no longer has to carry the historical explanation on its own.
This makes GifCities useful for a reader interested in ordinary expression rather than only famous destinations. A decorative image can be considered alongside the writing and arrangement around it. The question becomes what the page was trying to communicate. Enthusiasm, welcome, allegiance and a desire to fill a corner are all possible readings to investigate, rather than labels to attach in advance.
Careful looking also leaves room for uncertainty. Finding an image on a page does not by itself establish who drew it. Placement, authorship and preservation are separate questions. A responsible reading can describe the visible relationship without inventing a creator biography. There is enough interest in the page itself. An attractive pixel object does not need an unsupported origin story added to its luggage.
Two ways to describe the same picture
The current GifCities About page documents a 2025 update that added a semantic search index alongside the earlier index built from filenames and directory paths. This matters when reading older descriptions of the site: an account of the original search is not a complete guide to the current version.
The Special Search instructions identify three choices: semantic, URL text and both. Semantic search matches a description of the image; URL text searches the words present in its address and filename. Both is the documented default. These are documented behaviours, not results of a comparative accuracy test conducted for this article.
The two approaches invite different questions. A descriptive query asks what an image seems to show. An address-based query asks what words occur in the text associated with its location. Those questions can coincide, but they need not. Treating them as interchangeable would make it harder to understand why changing the search mode might be worth an experiment.
An illustrative exercise would begin with the same subject in each mode. A visitor interested in a computer drawing could try a description first, then a simple word in URL text mode. The useful observation would be what changed, rather than whether one set of results could be declared the winner. No particular result is promised by that example.
The distinction also shows up in the homepage links. During the source check, the homepage sent its drawing-of-cat suggestion to Special Search with semantic mode selected. The snowglobe suggestion used the ordinary search route. That is an inspectable interface detail, rather than a test showing that either mode is more accurate.
Search notes should retain that difference. A useful record contains the words entered, the mode selected and any dimensional conditions. Those pieces explain what the visitor asked the service to retrieve. A bare screenshot might show an interesting image, but it would leave the question behind the results unclear. A little documentation makes a return visit more intelligible.
The semantic mode also calls for care in historical interpretation. A modern description of a picture is not necessarily wording supplied by the original page maker. The filename route supplies a different clue. Neither should be promoted automatically into a statement about authorship, intention or the original visitor's experience. Search helps locate something worth examining; the explanation still needs evidence.
When the dimensions are the question
According to GifCities Special Search, width and height filters perform exact matches. Its own example is a width of 88 and a height of 31. These fields are useful when the reader's curiosity concerns a shape or a particular amount of space, rather than only an object named in a query.
Exact matching is a practical detail with consequences for how an empty search should be read. A requested size is a condition, not a general preference for something roughly similar. The absence of a result under a combination of conditions would not establish that no related picture exists anywhere in the archive. The visitor has asked a narrower question than that.
One proposed exercise is to hold the dimensions steady while changing the subject. Another is to keep the subject and remove the size conditions. These are two different comparisons. The first concentrates attention on how different things might fit a fixed space. The second asks whether a dimensional restriction was excluding otherwise interesting candidates. Neither requires a claim about the collection's completeness.
There is a pleasing seriousness to specifying the exact size of something frivolous. The object might be decorative, but the rectangle is not vague. That combination suits the site. It allows a visitor to approach a tiny animation with the care usually reserved for something much larger, without requiring the animation to become more important than it is.
Dimensions also provide a useful editorial restraint. A writer looking at a small object can describe its bounds before making claims about what it meant. What occupies the space? What seems to move? What remains still? Those questions encourage an account of the object in front of the reader. Broad declarations about the old web can wait until there is evidence for them.
For a browsing session, the fields need not become a technical project. A single comparison is enough to make the control intelligible. The reward is a better sense of what the search is being asked to find. Exact sizes are most useful as deliberate constraints. Entering them simply because the boxes exist would turn a playful visit into unnecessary paperwork.
Following a picture back to a page
The About page says search results link back to the original GeoCities pages in the archive. It also identifies the GeoCities Closing Crawl as the project's source collection. This article verifies that documentation, rather than claiming that every destination has been opened or that every archived page renders completely.
A live retrieval of the snowglobe results page returned page-one-of-six navigation and linked images pointing to timestamped Wayback Machine addresses. The HTML was inspected on 12 November 2026. That confirms a working result response and archive links for this query at that time; it does not confirm the completeness of the archive destinations.
The first linked destination in that response used a July 2009 capture timestamp. Other visible links used later 2009 timestamps and different GeoCities addresses. Those are observations of the destination URLs, not findings about what the pages contain. The address can tell a reader where the next step leads without proving that every asset at that destination will load.
The page label is similarly bounded evidence. Six pages for one query does not provide a count of unique images in the entire collection. Nor does a page-size parameter prove that every page contains that many results. The retrieved response supports a modest conclusion: this particular search supplied a navigable set of links. Larger totals belong with the project's own dated documentation.
A useful visit can follow three questions: what attracted attention in the image, what surrounded it on the archived page, and whether that surrounding material changes the first impression. This is a proposed method of reading, not a claim about a tested route through the interface. Its purpose is to make the linked page more than a source of additional ornaments.
The first question is visual and limited. Perhaps the interest lies in the shape, a motion or a particularly earnest decorative idea. The second question is relational. The page supplies neighbours. The third question gives the visit a direction: the reader can revise the initial interpretation when more context appears. A little moving picture can support that sequence without requiring a grand theory.
There is also value in preserving the distinction between an archived page and a present-day account of it. An old page can supply evidence of what appears in that capture. It cannot, on its own, explain everything that happened before or after. An archive visit benefits from patient descriptions of what is visible, with larger conclusions held to a correspondingly larger evidential standard.
This is where the apparent triviality of GifCities becomes useful. A butterfly is a manageable starting point. The reader can practise the difference between noticing, interpreting and knowing without needing a major historical event as the subject. The smallness keeps the exercise approachable. It does not remove the need to distinguish a pleasing guess from something supported by a source.
It also suggests a courteous way to write about personal pages. Their decoration can be described without assuming that the maker was naive, ironic or representative of everyone else. A particular arrangement deserves to remain particular for a while. The archive is more interesting when its inhabitants are allowed different intentions, including intentions that a later visitor cannot confidently recover.
A postcard made from the collection
GifCities GifGram adds a second way to spend time with the pictures. Its opening instructions describe an old-web e-card maker: arrange GIFs in a collage, add text, then save the composition to obtain a unique URL for sharing. This article inspected those instructions and controls; it did not save a card or send a message.
The distinction matters because the maker is an activity with its own outcome. Search asks for an object. A card asks for a relationship between objects, words and a recipient. The decorative image becomes part of a new arrangement. That is an editorial reading of the task described by the maker, rather than a claim that every resulting card has a particular effect.
The GifGram interface exposes controls for adding a GIF, adding text, choosing a background and saving. Its on-screen keyboard guide lists undo, clearing the cursor, scaling, rotation and tilt. These visible instructions establish what the maker advertises. They do not establish that every shortcut works on every device or that the saved link will remain available indefinitely.
A small proposed card could have a single subject and a short line of text. That would make arrangement the question: which part should draw attention, and how much movement suits the message? The exercise needs no claim about historical authenticity. A present-day collage made from archived material can be understood as a new composition, rather than a recovered original page.
The save step deserves its own boundary. The maker says it supplies a unique URL, which is a way of sharing a composition. A unique address is not, by itself, a documented privacy guarantee. This field guide does not recommend putting confidential material into a card. The verified feature is the described link-based output, not a tested account of its access controls.
A visitor can also decline the making part. The collection remains interesting as a search destination, and no composed card is needed to justify the visit. GifGram is worth distinguishing because it changes the role of the pictures: a search result can become material for a message. It does not make every search a preliminary step towards producing something.
The connection to the Boring Websites network is a matter of attention. Today's Rock is a fitting next stop for a reader inclined towards a small subject. GifCities offers a different smallness, with an archived context behind the objects. This is an editorial association, rather than a claim that the services share ownership.
The wider Boring Websites network gives this field guide its home. GifCities belongs in its orbit because a modest premise can sustain a proper look. A search box, an ornament and an archived page are enough for a visit. A visitor may arrive seeking a butterfly and leave interested in the page that gave it somewhere to flap.
FAQ
What is GifCities?
GifCities is the Internet Archive's GeoCities animated GIF search project. This field guide treats its images as an entrance to archived personal pages, rather than as a survey of every animation on the web.
Does it still search filenames?
Yes. Special Search documents URL text mode alongside semantic mode and the default combination of both. The modes search different kinds of descriptions, so their results should not be assumed identical.
Can searches specify an image size?
The Special Search documentation supports exact width and height matches, with 88 by 31 as its example. Those are exact conditions rather than approximate size preferences.
Were all the archived pages tested for this article?
No. The About page documents links from results to archived GeoCities pages. This article verifies the published explanation and the search form, without claiming a complete test of the collection or its linked destinations.